Low-complexity inverse transform and sub-sampling

1. A method of inverse transform and sub-sampling having low computational complexity, comprising: separating input signals into high-frequency signals and low-frequency signals; expressing said high-frequency signals as a high frequency matrix in spatial domain, and expressing said low-frequency signals as a low frequency matrix in spatial domain; performing compensation computations according to H 1 −1 ×H 2 ×Y 3 +Y 2 ×K 2 ×K 1 −1 ×H 1 −1 ×H 2 ×Y 4 ×K 2 ×K 1 −1 , wherein H 1 −1 is the inversed first portion of the combined matrix from the down-scale matrix and the transposed inverse transform matrix, H 2 is the second portion of the combined matrix from the down-scale matrix and the transposed inverse transform matrix, Y 3 is the left-bottom portion of the frequency-domain coefficient matrix, Y 2 is the right-top portion of the frequency-domain coefficient matrix, Y 4 is the right-bottom portion of the frequency-domain coefficient matrix, K 2 is the second portion of the combined matrix from the inverse transform matrix and the transposed down-scale matrix, K 1 −1 is the inversed first portion of the combined matrix from the inverse transform matrix and the transposed down-scale matrix; obtaining compensation values for non-zero high-frequency signals; adding separated low-frequency signals with said compensation values; selecting a dimension of an inverse transform matrix depending on actual sub-sampling requirements for said inverse transform and said sub-sampling; said dimension of said inverse transform matrix being less than that of an original inverse transform matrix; integrating a matrix of said inverse transform with said matrix of said sub-sampling in cooperation with compensated frequency-domain coefficients, in proceeding with said inverse transform and said sub-sampling of said matrix; separating said input signals into said high-frequency signals and said low-frequency signals being based on ways of sub-sampling; determining the dimension of the inverse transform matrix according to reduction ratios of a frame length and width; utilizing an original frame size L 1 ×W 1 in reducing a frame size into L 2 ×W 2 ; reducing said dimension of said inverse transform matrix through reduction ratios of L 2 /L 1 and W 2 /W 1 ; realizing a frame-size reduction transform, and determining a plurality of said low-frequency signals; and performing an inverse transform and a sub-sampling for a result of additions.

2. The method of inverse transform and sub-sampling having low computational complexity as claimed in claim 1 , wherein performing said compensation computations and obtaining said compensation values for said non-zero high-frequency signals, said compensation computations are used to produce said compensation values by multiplying items of a table with said non-zero high-frequency signals.

3. The method of inverse transform and sub-sampling having low computational complexity as claimed in claim 2 , wherein said compensation computations are performed by a compensation module, and said compensation module is provided with a built-in table, or said compensation module produces said table through computations.

4. The method of inverse transform and sub-sampling having low computational complexity as claimed in claim 1 , wherein in performing said compensation computations for said non-zero high-frequency signals, an end-of-block point is used, so as to perform said compensation computations for non-zero high-frequency coefficients before said end-of-block point, hereby reducing a plurality of determinations of said non-zero high-frequency signals.

5. The method of inverse transform and sub-sampling having low computational complexity as claimed in claim 1 , wherein said inverse transform is one of following: an inverse discrete cosine transform, an inverse discrete sine transform, an inverse discrete Fourier transform, an inverse discrete wavelet transform, and various other inverse transforms.

6. The method of inverse transform and sub-sampling having low computational complexity as claimed in claim 1 , wherein said sub-sampling includes filtering of data or not performing any pre-processing, and said sub-sampling is classified into: a regular interval sub-sampling, a non-regular interval sub-sampling, and various other sub-samplings.